Several methods for nodal regionalization using interaction matrices such as information, passengers, commodities and money, have been developed since around 1960s, including Graph theory method (1) (Nystuen & Dacey, 1961), Direct connection method (Morikawa, 190'1), Principal component analysis method (Garrison & Marble, 1962), Factor analysis method (Goddard, 1970), Markov chain method (Brown & Holmes, 1971), Combination method of graph theory and gravity model (Rouget 1972) Graph theory method (2) (Slater 1974) Direct cluster analysis method (Clayton, 1974) and Functional distance method (Keane, 1978). Of these methods, Nystuen-Dacey model has been most widely used for empirical studies (Tinkler, 1979). Objectives of this paper are, first, to provide a brief review of various modifications of the Nystuen-Dacey model (Table 1) and, second, to test empirically an improved model by the author. This improved mode_1 contains six main modifications, including: (1) A study area is regarded as an open system, by considering also the interactions with outside the study area, (2) Nodal hierarchies are determined by the rank size distribution of total flows each node generates, (3) Indirect effects of n-steps are evaluated by employing equation (6), (4) Directionality of nodal flows is calculated by employing equation (8), (5) The nth largest nodal flows as well as the largest are evaluated to reduce the information loss of the nodal structures, by employing the method by Holmes & Haggett (1977), and (6) Dominant-subordinate relationship is measured by the method of Rouget (1972) (equation (7)). This model was applied to Yamanashi Prefecture and traffic flows were used as the interaction data. The result of the analysis (Figure 3, Figure 4 and Figure 5) was highly fitted with the previous empirical research, and the usefulness of the model was examined.
抄録全体を表示